PDF/A and PDF/UA are the two ISO PDF standards most often confused, partly because both add structure to a file and partly because "PDF/A-1a" sounds like it should mean accessible. This guide separates the two, explains the levels, and shows how to produce a file that satisfies an archive and a screen reader at the same time.
What PDF/A guarantees
PDF/A, the ISO 19005 series, is targeted at reliable long-term preservation of documents with text, images, vector graphics, and metadata. It defines a subset of PDF that can be reproduced faithfully years later: fonts must be embedded, colour must be specified device-independently, encryption is not allowed, and features that depend on external resources or executable content are excluded. The parts build on successive PDF versions, and none replaces the earlier ones; PDF/A-1 files remain valid archival files.
| Part | Base PDF version | Levels | Notes |
|---|---|---|---|
| PDF/A-1 (2005) | PDF 1.4 | a, b | b preserves appearance; a adds tags and Unicode text |
| PDF/A-2 (2011) | PDF 1.7 | a, b, u | u adds Unicode text without requiring tags |
| PDF/A-3 (2012) | PDF 1.7 | a, b, u | Allows arbitrary embedded files |
| PDF/A-4 (2020) | PDF 2.0 | none; f and e variants | Tags optional; f allows embedded files, e targets engineering |
Why PDF/A-1a is not accessibility
Level a ("a" for accessible, in the original naming) requires structure information and Unicode text so that the logical structure and reading order survive. That is genuinely useful, but PDFlib's summary of the standards is precise about the limit: "PDF/A-1a/2a/3a require only the mere presence of tags, but don't go into detail regarding the nature and use of PDF tags." A PDF/A-2a file can have every paragraph tagged as a paragraph, no headings, no table headers, and no alternative text, and still conform. PDF/A-4 makes the point structurally by dropping the a/b/u levels; it encourages tagging for reuse and accessibility but does not require it.
What PDF/UA guarantees
PDF/UA, the ISO 14289 series, is the accessibility standard. It requires that all meaningful content be tagged with semantically appropriate structure, that decorative content be marked as artifacts, that figures carry alternatives, that tables and lists use the right structure, that the document declare its language and title, and that security settings not block assistive technology. Its test model, the Matterhorn Protocol, lists 136 ways to fail. PDF/UA-1 applies to PDF 1.7 files and PDF/UA-2 to PDF 2.0 files.
Both at once
The two standards are compatible by design, and dual conformance is common in practice. The PDF Association publishes the Matterhorn Protocol 1.1 as a file that conforms to both PDF/UA-1 and PDF/A-2a and describes it as a reference-quality PDF/UA file. Prince, the HTML-to-PDF engine, documents combined profiles PDF/A-1a+PDF/UA-1, PDF/A-2a+PDF/UA-1, and PDF/A-3a+PDF/UA-1. In Microsoft Word, the Save As PDF options dialog lets you select both document structure tags and PDF/A compliance in one export. The recipe is the same everywhere: produce a correctly tagged file, then apply the archival constraints (embedded fonts, no encryption, output intent), and validate against both.
Courts, archives, and records offices
United States federal courts require documents filed in CM/ECF to be PDF, and the Judiciary has stated its intention to move to PDF/A for security and archiving; individual districts publish their own guidance, and the Eastern District of Oklahoma, for example, says PDF/A-1a is preferred and PDF/A-1b meets the minimum. None of that is an accessibility requirement. An organisation with both a records mandate and an accessibility mandate, which describes most public bodies under the ADA Title II rule and the European standard EN 301 549, needs to state both in its document policy, because a PDF/A pipeline will not produce accessible files on its own.
Practical guidance
- Set accessibility requirements at the source document, then choose the PDF/A part the archive needs; the export tools above support both in one step.
- Validate separately: an archival validator for PDF/A and a PDF/UA validator or checker for accessibility. The free PDF checker covers the machine-detectable accessibility side.
- Prefer XFA-free AcroForms if a form must be archived; XFA is not permitted in PDF/A and was dropped from PDF 2.0. See accessible PDF forms.
- For documents people actually read online, consider whether the archival copy and the reading copy should be the same file at all; a PDF/A master plus hosted accessible HTML serves both needs better than one compromise file.
Frequently asked questions
Is a PDF/A file accessible?
Not necessarily. PDF/A (ISO 19005) is an archival standard. Its level a variants (PDF/A-1a, 2a, 3a) require tags and Unicode text but only their presence, not their correctness, and PDF/A-4 makes tags optional. Accessibility is the job of PDF/UA (ISO 14289), which requires semantically correct structure, alternatives, language, title, and security settings that allow assistive technology. A file can conform to both.
Can a PDF be both PDF/A and PDF/UA?
Yes. The PDF Association publishes the Matterhorn Protocol as a file conforming to PDF/UA-1 and PDF/A-2a, and tools such as Prince offer combined PDF/A-2a+PDF/UA-1 output. The file must be correctly tagged, use embedded fonts and device-independent colour, and must not be encrypted, since PDF/A forbids encryption.
Which PDF/A level should I use if I also want accessibility?
For PDF 1.7 files, PDF/A-2a or PDF/A-3a combined with PDF/UA-1, since level a requires the tag structure PDF/UA builds on. For PDF 2.0 files, PDF/A-4 with tagging enabled combined with PDF/UA-2. In every case validate the accessibility side separately; the PDF/A level alone does not check it.
Sources
The statements above rest on the primary sources below. Where a source is a law or standard, the version and date named in the text are the ones checked on the review date. This guide is general information, not legal advice.
- PDFlib knowledge base: The PDF/A standards . Parts, base versions, conformance levels, PDF/A-4 changes, and the quoted statement on level a.
- PDF Association: What is PDF/A?
- PDF Association, April 22, 2021: Matterhorn Protocol 1.1 is now available . The protocol PDF conforms to both PDF/UA-1 and PDF/A-2a.
- Prince documentation: PDF output and profiles . Combined PDF/A plus PDF/UA-1 profiles.
- United States Courts: Electronic Filing (CM/ECF)
- U.S. District Court, Eastern District of Oklahoma: PDF/A Frequently Asked Questions
- Datalogics: XFA forms deprecated, what it means and what to do . XFA is not permitted in PDF/A.
Keep reading
PDF/UA versus WCAG
What each standard measures, where automated validation stops, and what a defensible claim requires.
Read the guide →The Matterhorn Protocol explained
What the Matterhorn Protocol is, how its 31 checkpoints and 136 failure conditions test PDF/UA-1, which ones a machine can decide, and how to read a checker report built on it.
Read the guide →Accessible PDF forms: AcroForm vs XFA and WCAG
Why XFA forms fail accessibility and archiving, how AcroForm fields get an accessible name through tooltips, the WCAG PDF techniques for forms, and when an HTML form is the better answer.
Read the guide →